Data preparation
1. Data preparation
SPEAKER: Let's start with the first stage of ML workflow, data preparation. During this stage, you must upload data and then prepare it for model training with feature engineering. The data can come from Cloud Storage, BigQuery, or even your local machine. AutoML now mainly supports tabular data by solving different types of problems called objectives. For tabular data, you can train the model to solve regression, classification, or forecasting problems. Forecasting is vital to many industries like retail. To learn more about how to build a forecasting model, please check the course titled Introduction to Vertex Forecasting and Time Series in Practice in the reading list. After the data is uploaded to AutoML, the next step is preparing it for model training with feature engineering. Imagine you're preparing a meal. Your data is like your ingredients, such as carrots, onions, and tomatoes. Before you start cooking, you'll need to peel the carrots, chop the onions, and rinse the tomatoes. This is what feature engineering is like. The data must be processed before the model starts training. A feature refers to a factor that contributes to the prediction. It's like an independent variable in statistics or a column in a table. Preparing features can be both challenging and tedious. To help, Vertex AI provides a service called Vertex AI Feature Store, which is a centralized repository to manage, serve, and share features. It aggregates the features from different sources in BigQuery and makes them available for both real-time, often called online, and batch, often called offline serving. Vertex AI automates the feature aggregation to scale the process with low latency. Additionally, Vertex AI Feature Store is ready for generative AI. It can manage and serve embeddings, which are the data sources in Gen AI. It also supports retrieving similar items in real time, ensuring low latency. The workflow of serving real-time, online features with Vertex AI Feature Store can be summarized as follows. One-- prepare the data source in BigQuery. Two-- optional, register the data sources by creating feature groups and features. Three-- configure the connection by creating a feature view to define which features to copy from your data source into the online store for real-time serving. Four-- serve the latest feature values online from a feature view. So what are the benefits of Vertex AI Feature Store? First, features are shareable for training and serving. They are managed and served from a central repository, maintaining consistency across your organization. Second, features are reusable. This helps to save time and reduces duplicated efforts. Third, features are scalable. They automatically scale to provide low-latency serving, so you can focus on developing the logic to create them without worrying about deployment. And fourth, features are easy to use. Vertex AI Feature Store is built on an easy-to-navigate user interface.2. Let's practice!
Create Your Free Account
or
By continuing, you accept our Terms of Use, our Privacy Policy and that your data is stored in the USA.